Position Uncertainty and Localization in Distributed Networks

Abstract

A tremendous development of wireless communications technology and an intellectualization of everyday products are accelerating an implementation of the Internet of Things (IoT). The implementation of IoT has introduced technologies such as remote sensing, automated home appliance control, health care. One of the most promising applications for the IoT is a location based service (LBS) that accurately locates each device and uses that information. Many applications are being developed, such as searching for missing animals, autonomous driving, smart factories, platooning, automatic parking, customized shopping malls, and so on. <br/> Nowadays, technologies such as GPS, NB-IoT, LoRa, and Wi-Fi-based positioning system (WPS) are being used to provide such location-based services. However, the satisfactory positioning performance has not yet been achieved. GPS can not be used indoors and requires additional GPS modules. NB-IoT and LoRa are merely location 'notifications' method to effectively send GPS signals to communications network. Although WPS can estimate the position in indoor environment, it is difficult to analyze a signal due to the fluctuation of a RSSI.<br/> The problem of unsatisfactory location estimation performance of existing technologies is due to the lack of available information which is used for position estimation. The existing technology does not use signals from a large number of nearby devices, but uses only the signals provided by a base station or an access point. It is essential that each device communicates with the neighboring device for effective location estimation. With the opening of the 5G wireless communication, fortunately, technical support for massive connection is becoming possible and eventually it is expected that the IoT network will be configured as a mesh network with all devices connected to each other through device-to-device communications.<br/> In this dissertation, I propose location estimation techniques that can be used in distributed networks. In the distributed position estimation method, the position error of the device is inevitably accumulated in the position estimation error of the nearby device. It is called an error propagation phenomena. Therefore, I used Dilution of Precision (DOP) to perform quantitative analysis of the location inaccuracies and analyze the position error reduction obtained by considering position inaccuracy. <br/> Next, I proposed a distributed position estimation algorithm that takes into account the quantified position inaccuracies of neighbor devices. In this algorithm, I uses a second-order polynomial function to determine the final position of the device, which is determined as the balance point between the distance measurement from neighbor devices and the position inaccuracy of neighboring devices. By using this method, it is possible to prevent the degradation of the system performance by the unpredictable large error. In addition, the position estimation of the target device and the position correction of neighboring devices can be performed at the same time. At this time, since the position correction of neighboring devices is determined deterministically by the position of the target device, a computational complexity is reduced.